regex filter on non-string columns raises an exception

[jseabold]

I'm not really sure what should be done here. At least a better error message I suppose. Perhaps just skip non-string columns or maybe even try to turn them into strings? E.g., make match("\d", 123) work? Obviously could run into some problems for things that can't be converted (or do "unexpected" things) for turning into string/unicode.

>>> pd.version.version
>>> '0.12.0-1149-g141e93a'

(if columns are all numerical also bombs with different msg)
>>> pd.DataFrame(np.random.random((3,2)), columns=['STRING', 123]).filter(regex='STRING')
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-427-e5acd93a79d6> in <module>()
----> 1 pd.DataFrame(np.random.random((3,2)), columns=['STRING', 123]).filter(regex='STRING')

/usr/local/lib/python2.7/dist-packages/pandas-0.12.0_1149_g141e93a-py2.7-linux-x86_64.egg/pandas/core/generic.pyc in filter(self, items, like, regex, axis)
   1522             matcher = re.compile(regex)
   1523             return self.select(lambda x: matcher.search(x) is not None,
-> 1524                                axis=axis_name)
   1525         else:
   1526             raise TypeError('Must pass either `items`, `like`, or `regex`')

/usr/local/lib/python2.7/dist-packages/pandas-0.12.0_1149_g141e93a-py2.7-linux-x86_64.egg/pandas/core/generic.pyc in select(self, crit, axis)
   1116         if len(axis_values) > 0:
   1117             new_axis = axis_values[
-> 1118                 np.asarray([bool(crit(label)) for label in axis_values])]
   1119         else:
   1120             new_axis = axis_values

/usr/local/lib/python2.7/dist-packages/pandas-0.12.0_1149_g141e93a-py2.7-linux-x86_64.egg/pandas/core/generic.pyc in <lambda>(x)
   1521         elif regex:
   1522             matcher = re.compile(regex)
-> 1523             return self.select(lambda x: matcher.search(x) is not None,
   1524                                axis=axis_name)
   1525         else:

TypeError: expected string or buffer

[jreback]

yep...needs better error msg

[dylancis]

specially when dealing with multi-level index. We need to drop of the level to make it single index before be able to use filter method. Perhaps we could add a 'level' argument to the filter method so we can specify a level to look-at.

[mroeschke]

Looks to work correctly on master. Could use a test.

In [5]: pd.__version__
Out[5]: '0.26.0.dev0+684.g953757a3e'

In [6]: >>> pd.DataFrame(np.random.random((3,2)), columns=['STRING', 123]).filter(regex='STRING')
Out[6]:
     STRING
0  0.904452
1  0.546133
2  0.691725